A procurement committee asks three companies for a system to manage purchases across several branches. One bidder quotes for an application. Another quotes for discovery and a limited pilot. A third includes data cleanup, branch training and ongoing support. The committee arranges the prices in ascending order and prepares to recommend the cheapest. Then a branch manager asks which proposal covers a delivery received against the wrong purchase order. Nobody can answer.
If you are searching for the best software development company in Kenya, the useful question is which company can deliver the outcome your organisation needs under conditions you can sustain. A general ranking cannot settle that question. A hospital, an equipment distributor and a professional association may each choose a different team for sensible reasons. The word best earns its meaning only when you define the work and the evidence that will prove it.
Consider a fictional company buying a procurement application for Nairobi, Nakuru and Kisumu branches. It wants head office to see which requests have been approved, which goods have arrived and which supplier invoices still need checking. The example below shows how to build a shortlist, compare bids, moderate a scoring meeting and turn the winning proposal into a pilot with agreed evidence.
Write down what a wrong choice would look like
Imagine the system six months after an unsuccessful launch. Branch staff still send requests through personal messages because the application does not recognise urgent purchases. Head office sees a polished report but cannot distinguish a received order from a supplier invoice. The original developers understand a crucial integration, and no one else can repair it. These are specific failures. They can become selection criteria that matter more than a generic list of desirable features.
Invite each stakeholder to describe one failure they would find unacceptable. For stores it might be stock recorded before physical receipt. For finance it might be paying an invoice twice. For a branch manager it might be losing authority to approve a small urgent repair when head office is closed. Ask them to explain the consequence and how they would recognise the failure. It is easier to test an observable consequence than a broad demand for quality.
Then describe a good outcome in equally concrete terms. A purchase request becomes an approved order, goods arrive in two deliveries and the invoice is checked against what actually arrived. A reviewer can identify who approved the exception and why. The branch can follow up outstanding quantities without calling accounts. The supplier receives a clear answer about the disputed line. These observations give the committee a shared basis for evaluating the proposed workflow.
Use the failure statements to distinguish essential conditions from preferences. If the buyer must retain access to its data and the candidate cannot explain how, that may justify exclusion before ranking. A different colour scheme or a slightly different meeting schedule may be negotiable. Keeping these categories separate prevents a high score for presentation from offsetting an unacceptable weakness in ownership or operational control.
Design a comparison that treats bidders fairly
Give every bidder the same documents, assumptions and response deadline. Use sample data with confidential information removed. Describe the current process and the outcome you seek, including systems that must remain in place. State whether bidders may propose configuration of an existing product, a custom extension or a new build. Otherwise one team may price a replacement for everything while another prices only the gaps.
Publish the evaluation method internally before reading proposals. Decide which people will score technical delivery, operational fit, commercial terms and support. Make sure branch staff can comment on the parts they will use. The finance director may be able to judge invoice reconciliation, while a receiving clerk can identify whether a proposed goods receipt would work when a vehicle arrives after normal hours. Both voices matter to the decision.
Give each candidate the same opportunity to clarify. If one asks whether branches can purchase independently during an outage, share the business answer with all competing teams where appropriate. A useful question from one bidder can improve every proposal. Preserve the bidder’s own confidential material while ensuring that candidates are working against the same buyer assumptions. This simple discipline prevents accidental comparisons between different scopes.
Specify what constitutes evidence. A screenshot shows that a screen exists. A test using your sample transaction shows something about behaviour. A proposal commitment identifies a future deliverable. A reference call gives another person’s account of previous delivery. These are different kinds of evidence with different limits. Ask reviewers to record which kind supports each judgement so confidence does not grow simply through repetition at the committee meeting.
Select a transaction that crosses real authority boundaries
Choose a transaction that begins at one branch and ends with a decision at head office. In the example, Nakuru requests thirty boxes of supplies. Head office approves twenty four. The supplier delivers twenty, then invoices twenty four. The branch expects another delivery, finance wants to know what can be matched and the supplier wants to know why part of the invoice is under review. A good demonstration can make all three positions visible.
Give the bidder the documents and ask for a walkthrough. A preliminary design exercise can be done on paper or with an existing test environment. Do not expect a vendor to build a production feature for free. Explain what you want to learn and limit the exercise. If the team needs deeper paid discovery to address critical unknowns, ask it to describe the outputs of that discovery and the decision those outputs will enable.
Observe where the bidder places authority. Can the receiving clerk alter the approved order to make it match the delivery? Can finance mark all goods received simply to clear an invoice? Who decides whether to accept the short delivery and close the remaining quantity? A strong response describes separate records for the request, approval, delivery and invoice while explaining how an authorised correction moves through the system.
Introduce a second delivery after the invoice is under review. Ask the branch clerk to record four boxes against the original order and finance to complete the match. The quantity and supplier references should reconcile without erasing the earlier short delivery. Ask for a report showing what was outstanding at the end of the first day. That final question distinguishes a current total from a history capable of explaining an earlier decision.
Build a scorecard that measures what you mean
Choose a small number of criteria and define each in ordinary language. Operational fit might mean that a branch can record partial receipt, a supervisor can authorise a substitution and finance can trace the final invoice. Delivery capability might mean that the proposed team can demonstrate a working slice, explain how it tests changes and show how open questions are resolved. Give each score a description so reviewers know what earns it.
Suppose the committee allocates thirty points to operational fit, twenty to delivery capability, twenty to support and continuity, fifteen to the commercial proposal and fifteen to adoption. These figures are an illustration, not a universal model. A buyer with round the clock operations might assign more weight to recovery. A small association with limited internal capacity might place greater emphasis on training and administrative simplicity. Set the weights before reading bids.
Define how missing evidence is treated. A candidate may honestly say it needs to inspect an external accounting interface before confirming a connection. That uncertainty is different from a confident claim unsupported by a demonstration or dependency check. Record the uncertainty, the potential impact and the proposed method of resolving it. The committee can then decide whether to commission discovery, impose a condition or exclude a proposal with an unacceptable gap.
Avoid turning the score into a false measurement. A difference of a point or two may reflect interpretation rather than a meaningful advantage. Look at the underlying evidence and discuss what would change the decision. If one bidder clearly handles branch adoption better and another has stronger integration experience, identify which risk threatens this project more. The scorecard organises judgement. It cannot remove the committee’s responsibility to exercise it.
Examine the people who will actually deliver
Ask who will attend branch discovery, who will make technical decisions and who will answer when a pilot transaction fails. A company may have a large team while assigning only a small part of it to your project. A smaller company may have strong direct access to its lead but a thinner continuity plan. Evaluate the proposed people, their availability and the way knowledge passes between them.
Use a scenario to test collaboration. The branch supervisor and head office disagree about urgent purchases. Ask the proposed business analyst or project lead to facilitate a short discussion. Watch how they draw out examples, state unresolved choices and identify who can decide. An impressive technical answer is less helpful if it settles an organisational dispute without authority. The team should know when to write a rule and when to bring a decision back to management.
Ask the technical lead to explain a difficult change to someone who does not write software. Suppose the approved quantity must be preserved after a partial delivery. What changes in the records, reports and tests? A clear answer will address the business effect and the work involved. If every explanation relies on unexplained terminology, branch staff may struggle to report defects or participate in acceptance testing once the project is underway.
Verify continuity through an example. If the lead leaves during the pilot, where would a replacement find the decision about partial receipts and the instructions for running its tests? Ask for a demonstration using nonconfidential material. A maintained decision record and reliable test instructions make continuity believable. A list of staff names alone cannot establish whether another person could safely take over the project.
Read commercial proposals on a common basis
Create a comparison with the work down the left and each bidder across the top. Include discovery, application changes, data migration, integration, devices, training, support and any recurring services. Mark items as included, excluded, dependent on investigation or supplied by the buyer. An empty cell needs a question. Do not quietly treat it as a zero cost merely because the price table lacks a corresponding line.
Set an operating scenario for the comparison. Use the same number of branches, expected volume of purchase documents, approximate attachment pattern and support hours for every candidate. Then vary one factor, such as opening a fourth branch. Ask what changes in setup effort, licence charges, training and ongoing support. You are looking for cost behaviour under realistic conditions, not a promise that the final bill can never change.
Examine buyer effort. One bidder may require finance to clean vendor records, while another includes assisted cleanup and a rehearsal. Either arrangement could work. The committee needs to know whether the finance team has the time and skills to perform its part. An external price that appears attractive can become a poor value if the buyer must repeatedly divert key staff from closing accounts to resolve migration problems.
Ask what each payment milestone buys. A useful discovery milestone can deliver agreed workflows, a dependency list and an implementation estimate. A pilot milestone can deliver working branch transactions reconciled by finance. If a milestone is described only as fifty percent development complete, ask for an outcome the buyer can inspect. Better definition supports both parties when deciding whether an invoice is due and what remains to be delivered.
Check references with questions a former client can answer
Ask for permission to speak with relevant previous clients. Respect confidentiality and avoid asking for proprietary details. A buyer can learn a great deal from the pattern of delivery. Did the supplier make assumptions visible? What happened when a branch discovered that an agreed workflow did not fit? Could the internal team operate the system after launch? Open questions invite a more useful account than asking whether everything went well.
Look for a reference with a similar operating challenge. The industry need not match exactly. A supplier who handled distributed approvals, partial deliveries and a difficult migration in another sector may have relevant experience. Ask what made the project comparable and which differences matter. A reference from a prestigious organisation is less useful if the development company supplied only a small unrelated component.
Ask about the people involved. The company under consideration may have changed its delivery team since the reference project. Learn whether the proposed lead participated and whether the methods described are still used by the people assigned to your work. Treat a past success as helpful evidence with limits. Follow up through the pilot on the specific capability your organisation needs today.
Record reference comments separately from committee interpretations. If a client says the vendor responded quickly during launch week, that supports a statement about one period of support. It does not automatically establish a particular response commitment for your contract. Ask the candidate for its proposed support arrangement and compare it with the reference experience. The distinction keeps useful testimony from becoming an unsupported promise.
Study how the provider plans for failure
Give the candidate an incident. A branch records a goods receipt, loses the connection and submits it again. Head office subsequently sees two receipts. Ask how the team would prevent the duplicate where possible, detect it if it occurred and repair the records without concealing the original sequence. A response that depends on a clerk deleting a row from a database is unlikely to satisfy finance’s need for an explanation.
Consider a wider outage. If branches cannot access the application during a morning delivery window, who authorises a temporary receipt and where is it recorded? Agree how temporary records will later enter the system and who checks that no goods were counted twice. Ask the candidate to describe what its support team does while branches operate the fallback. A practical answer will take account of both recovery and reconciliation.
Explore the backup plan by asking for a restoration exercise. A successful backup job does not prove that documents, attachments and configuration can be restored together. Ask what the provider will restore, where it will do so and how the buyer can verify a sample purchase from request to invoice. Establish who has access to the necessary accounts. A backup under an inaccessible account can be as unhelpful during an incident as no usable backup.
Ask how the supplier reports an incident after service is restored. The buyer needs an explanation of the cause, the records affected, any remaining uncertainty and the preventive work proposed. Avoid demanding certainty before the facts are known. A team that can state what it has verified and what remains under investigation gives management a better basis for decisions than a team that offers reassurance without evidence.
Moderate the scoring meeting carefully
Have reviewers score independently before discussing the bids. This reduces the chance that the most senior person’s preference becomes everyone else’s starting point. Ask reviewers to write the observation behind each score. The receiving clerk might note that one candidate preserved the outstanding four boxes after the first delivery. Finance might note that another candidate could show the approved amount but could not reproduce the end of day position.
Discuss disagreements openly. One reviewer may favour a candidate for its simple interface while another is concerned about a missing audit trail. Ask whether the interface can remain simple while the record retains the needed history. The disagreement may reveal a solvable design question or a fundamental mismatch. Invite clarification from the bidder where appropriate and share material changes in scope fairly across the competition.
Challenge scores that depend on confidence alone. If a provider says it has done many integrations, ask which evidence supports the particular integration in your scope. If a bidder shows a modest prototype that handles the difficult transaction, give the observed behaviour its proper weight. Reviewers should be able to explain why a candidate earned a point using something more precise than the feeling that its team seemed professional.
Document the decision and its conditions. If the winning candidate must confirm accounting access during discovery, say what happens if access is unavailable. If a lower scoring candidate remains suitable as an alternative, record the reason. A clear decision note helps future managers understand why the project was awarded and gives the selected supplier a precise account of the risks it is expected to resolve first.
Award a pilot that can disprove your assumptions
Make the pilot small enough to manage but broad enough to cross the important handover. A single branch request, head office approval, two deliveries and an invoice reconciliation can exercise more meaningful integration than a dozen isolated screens. Define the test data, responsible users and evidence needed for acceptance. Give staff time to rehearse the process and to report where the design forces them into a workaround.
Set conditions for expansion. The branch clerk must be able to distinguish ordered, received and outstanding quantities. Finance must be able to explain the supplier’s balance against the supporting records. A deputy must be able to approve a legitimate urgent request when the normal approver is unavailable. Agree how unsuccessful tests are corrected and repeated. The conditions should leave room for learning while giving the buyer a clear decision point.
Allow a respected supplier to disagree. If the pilot shows that the buyer’s original approval hierarchy delays urgent purchases without adding control, the developer should be able to raise the issue. The business decides its own policy, but evidence from real transactions can lead to a better policy. Keep the decision and the reason in the project record so the change is understood when other branches join.
At the end of the pilot, compare the actual experience with the selection scores. Which claims were confirmed? Which assumptions changed? Did the people promised in the proposal do the work? Can the internal team explain the process without the project lead in the room? This review converts procurement from a one time competition into an informed decision about whether to expand the relationship.
Work through a close result before awarding the contract
Imagine two shortlisted suppliers finishing within a few points. One showed the purchase cycle clearly but proposed a support arrangement that ends before branch receiving closes. The other offered wider support coverage but could not explain how finance would reproduce the position after the first partial receipt. The committee should resist treating these as interchangeable weaknesses. One concerns the time at which expertise is available. The other concerns whether the accounting evidence will exist at all.
Ask whether either weakness can be resolved by a concrete condition. The first bidder might offer an extended support window or demonstrate a reliable branch fallback for the late receiving period. The second might provide a short proof of historical reporting using the sample transaction. Request comparable clarification under the buyer’s procurement process. A condition should specify the evidence, the responsible person and the point at which the committee decides whether the answer is sufficient.
Recalculate only the criteria affected by new evidence. If the historical report works in a demonstration, adjust the assessment of that capability. Do not automatically improve the entire bid because the team responded promptly. Equally, a supplier that charges extra for wider support may still be a good choice if the buyer genuinely needs it. Update the commercial comparison and show the effect of the added service over the agreed operating period.
Explore the consequences if the evidence fails. If finance cannot reproduce the partial receipt position, could the pilot operate with an authorised external reconciliation, or would that violate a mandatory condition? If the late support window remains unavailable, can branches safely continue with the fallback? The answer depends on the business’s own controls. Write the answer before making an award so the committee does not quietly lower its standards once implementation has begun.
Ask an independent colleague to read the decision note. Give them the brief, the critical transaction, the scores and the unresolved conditions. Can they explain why the preferred bidder won and what would cause the buyer to reconsider? If the explanation depends on the committee chair’s recollection of a persuasive presentation, improve the record. A clear award decision protects the supplier too, because it defines what the buyer expects and where further discovery is still required.
Separate a project reference from a project claim
A bidder may describe a previous implementation across several branches. Before giving that claim weight, ask what the company actually delivered. It may have built the central application, configured one module, supplied developers to a larger team or supported the system after launch. Each role can provide relevant experience, but they are different. A company that accurately describes its contribution gives the buyer better evidence than one that takes credit for an entire programme.
Ask what can be shown without exposing another client’s business. A generic process diagram, a demonstration using fictional records or a description of an incident may be sufficient. The buyer does not need another organisation’s live credentials or confidential transactions. A supplier that respects its former client’s boundaries while explaining the delivery method is demonstrating how it may treat your own information later. Record the specific lesson that transfers to your project.
Match the previous experience to the proposed team. A company may have delivered its strongest work through people now assigned elsewhere. Find out who performed the relevant discovery, testing or migration and how their knowledge entered the company’s working methods. A colleague need not have worked on the exact earlier project to benefit from it, but they should be able to explain the relevant procedure and its limits in clear terms.
Give the pilot sponsor a decision the board can understand
At the end of discovery, the sponsor should be able to state the next investment in a few concrete sentences. The branch will raise one request, head office will approve a smaller quantity, stores will record two receipts and finance will reconcile the invoice. The pilot includes the named users, agreed records and a defined period of support. External accounting access remains a dependency with an owner and a review date. This is a more useful funding request than asking the board to approve general automation.
Include a reason the board might stop or narrow the project. If the accounting provider cannot supply the necessary access, the buyer might pilot the operational workflow with a controlled export, change the integration design or choose another approach. The sponsor should know which option remains acceptable and what each would cost. Defining a stopping condition does not weaken commitment. It protects the buyer from continuing a project after the evidence supporting its original selection has changed.
Finally, record who can sign off the pilot. The branch receiving clerk verifies receipt handling, finance verifies reconciliation and the sponsor makes the expansion decision. A signature from one senior person should not conceal a known operational failure at a branch. Collect the evidence from the people who performed the work and identify any open conditions. The board can then see exactly which capabilities are ready for wider use and which still require investment.
Use the same evidence to brief the other branches. An incoming manager should understand why the first branch received twenty boxes against an approved quantity of twenty four and where the remaining four appear. Ask the manager to explain the example before planning rollout. If the explanation depends on access to a private spreadsheet at head office, the pilot has not yet eliminated the dependency that procurement identified. The sponsor can narrow the next release or commission the missing reconciliation before promising a broader deployment. A reliable award decision gives every branch a practical way to challenge the proposed operating assumptions.
Take a defensible question to the market
The best software development company in Kenya for your project is the one whose proposed team, methods and commercial terms meet the demands you can demonstrate. Build the evidence from an actual purchase, a partial delivery and a supplier dispute. A candidate that helps the receiving clerk and finance tell the same truthful story has answered a question that testimonials and broad rankings cannot settle.
Before your next procurement meeting, put one flawed transaction beside your scorecard and ask which criterion would expose the problem. If none would, revise the scorecard while the stakes are still low. For related buyer guides on evaluating, testing and operating business applications, continue with the ZamaCore software blog.